Ai Poh Loh

Neural network modelling and control strategies for a pH process

Abstract The control of a pH process using neural networks is examined. The neural network as a universal approximator is used to good effect in this nonlinear problem, as is shown in the simulation results. In the modelling task, the dynamics of the process was carefully examined to determine a suitable structure for the net. In particular, a multilayer net consisting of two single hidden layers was constructed to reflect the Wiener model of the pH process. This led to much simpler training compared to similar modelling attempts by other researchers. For the control task, two schemes were simulated. In one approach, a net was used to deal with the static nonlinearity to achieve control over a wide working range. The dynamic controller used was the PID, with its parameters tuned on a relay auto-tuner. This control design was compared with the strong acid equivalent method. In the second approach, a direct model reference adaptive neural network control scheme was proposed. The training procedure uses the more efficient least squares algorithm developed by Loh and Fong.

Modeling and control of a pilot pH plant using genetic algorithm

The work described in this paper aims at exploring the use of computational intelligence (CI) techniques for designing a Wiener-model controller to perform pH control. First, genetic algorithm (GA) is utilized to identify the static inverse titration relationship of a weak-acid strong-base titration process. The resulting model of the inverse neutralization equation then serves as the component in a Wiener model controller that linearizes the pH process. As the bulk of the system non-linearity is cancelled by the inverse model, a setpoint-weighted Proportional plus Integral plus Derivative (PID) controller is used to generate the control signal. A multi-objective evolutionary algorithm (MOEA) is employed to evolve a pareto optimal set of PID parameters in order to achieve the conflicting goals of fast rise time with small overshoots. Experimental results obtained from a laboratory-scale acid-base titration process are then presented to demonstrate the feasibility of the design methodology.

Adaptive control of a class of nonlinear systems with convex/concave parameterization

Abstract This paper deals with adaptive control of a class of second-order nonlinear systems with a triangular structure and convex/concave parameterization. In Annaswamy et al. (Automatica 33(11) (1998) 1975 –1995) it was shown that nonlinearly parameterized systems that satisfy certain matching conditions can be adaptively controlled in a stable manner. In this paper, we relax these matching conditions and include additional dynamics between the nonlinearities and the control input. Global boundedness and convergence to within a desired precision e is established. No overparameterization of the adaptive controller is required.

Brief Adaptive estimation of discrete-time systems with nonlinear parameterization

This paper concerns adaptive estimation of dynamic systems which are nonlinearly parameterized. A majority of adaptive algorithms employ a gradient approach to determine the direction of adjustment, which ensures stable estimation when parameters occur linearly. These algorithms, however, do not suffice for estimation in systems with nonlinear parameterization. We introduce in this paper a new algorithm for such systems and show that it leads to globally stable estimation by employing a different regression vector and selecting a suitable step size. Both concave/convex parameterizations as well as general nonlinear parameterizations are considered. Stable estimation in the presence of both nonlinear parameters and linear parameters which may appear multiplicatively is established. For the case of concave/convex parameterizations, parameter convergence is shown to result under certain conditions of persistent excitation.

Model-based contextual policy search for data-efficient generalization of robot skills

In robotics, lower-level controllers are typically used to make the robot solve a specific task in a fixed context. For example, the lower-level controller can encode a hitting movement while the context defines the target coordinates to hit. However, in many learning problems the context may change between task executions. To adapt the policy to a new context, we utilize a hierarchical approach by learning an upper-level policy that generalizes the lower-level controllers to new contexts. A common approach to learn such upper-level policies is to use policy search. However, the majority of current contextual policy search approaches are model-free and require a high number of interactions with the robot and its environment. Model-based approaches are known to significantly reduce the amount of robot experiments, however, current model-based techniques cannot be applied straightforwardly to the problem of learning contextual upper-level policies. They rely on specific parametrizations of the policy and the reward function, which are often unrealistic in the contextual policy search formulation. In this paper, we propose a novel model-based contextual policy search algorithm that is able to generalize lower-level controllers, and is data-efficient. Our approach is based on learned probabilistic forward models and information theoretic policy search. Unlike current algorithms, our method does not require any assumption on the parametrization of the policy or the reward function. We show on complex simulated robotic tasks and in a real robot experiment that the proposed learning framework speeds up the learning process by up to two orders of magnitude in comparison to existing methods, while learning high quality policies.

Describing function matrix for multivariable systems and its use in multiloop PI design

Abstract The describing function analysis for single-input single-output systems (SISO) is extended to multivariable plants under multiloop relay feedback control. Both necessary and sufficient conditions for limit cycle oscillations are given in terms of eigenvalues of the open loop system. The stability of such limit cycles is also established. Finally, the use of the describing function matrix in the design of multiloop PI controllers is demonstrated through some simulation examples.

Applications of Adaptive Controllers to Systems With Nonlinear Parametrization

Nonlinear parametrizations occur in dynamic models of several complex engineering problems. The theory of adaptive estimation and control has been applicable, by and large, to problems where parameters appear linearly. We have recently developed an adaptive controller that is capable of estimating parameters that appear nonlinearly in dynamic systems in a stable manner. In this paper, we present this algorithm and its applicability to two problems, temperature regulation in chemical reactors and precise positioning using magnetic bearings both of which contain nonlinear parametrizations. It is shown in both problems that the proposed controller leads to a significantly better performance than those based on linear parametrizations or linearized dynamics.

Integrated bake/chill module with in situ temperature measurement for photoresist processing

Thermal processing of photoresist are critical steps in the microlithography sequence. The postexpose bake (PEB) steps for current DUV chemically amplified resists is especially sensitive to temperature variations. The problem is complicated with increasing wafer size and decreasing feature size. Conventional thermal systems are no longer able to meet these stringent requirements. The reason is that the large thermal mass of conventional hot plates prevents rapid movements in substrate temperature to compensate for real-time errors during transients. The implementation of advanced control systems with conventional technology cannot overcome the inherent operating limitation. An integrated bake/chill module with in situ temperature measurement capability has been developed for the baking of 300-mm silicon wafers. The system provides in situ sensing of the substrate temperature. Real-time closed-loop control of the substrate temperature is thus possible as oppose to conventional open-loop control of the substrate temperature. Experimental results are provided to demonstrate a complete thermal cycle.

Adaptive control of discrete time systems with concave/convex parametrizations

This note considers the adaptive control of a class of nonlinear discrete time system with concave/convex parametrizations. The solutions involved two tuning functions which are determined by a minmax optimization approach much like the continuous time counterparts found in the literature. Direct extension from the continuous time case do not work very well due to the premature termination of the adaptive algorithm before zero tracking error can be achieved. In this note, this problem is solved. The proposed algorithm is shown to be stable and achieves zero tracking error in steady state.

ROBUST HUMAN DETECTION AND IDENTIFICATION BY USING STEREO AND THERMAL IMAGES IN HUMAN ROBOT INTERACTION

In this paper, robust human detection is investigated by fusing the stereo and infrared thermal images for effective interaction between humans and socially interactive robots. A scale-adaptive filter is first designed for the stereo vision system to detect human candidates. To eliminate the difficulty of the vision system in distinguishing human beings from human-like objects, the infrared thermal image is used to solve the ambiguity and reduce the illumination effect. Experimental results show that the fusion of these two types of images gives an improved vision system for robust human detection and identification, which is the most important and essential component of human robot interaction.